摘要
为解决通过人工及温度阈值分割评估蛋鸡羽毛覆盖度准确度不够的问题,提出基于热红外图像和彩色图像计算羽毛覆盖度的方法。利用Otsu算法结合不同的颜色模型提取鸡体目标及背部羽毛覆盖完好区域,通过计算面积比值,获取背部羽毛覆盖度。针对热红外图像目标鸡体边缘模糊导致分割效果较差的问题,提出基于分量R、B灰度直方图的自适应性热红外图像增强方法,该方法突出了前景与背景的差异,增强了边缘的清晰度。实验结果表明,热红外图像鸡体目标及羽毛覆盖完好区域的分割准确度达到97.18%和96.86%,彩色图像分割准确度达到99.58%和97.86%。对于羽毛覆盖度,热红外图像的计算结果接近羽毛覆盖度真实值。对同一批蛋鸡计算羽毛覆盖度,热红外图像比彩色图像的计算结果最少时高2.01个百分点,最大时高30.29个百分点。通过分析原因,发现蛋鸡剐蹭鸡笼等行为露出的白色羽毛根部,是彩色图像计算覆盖度结果出现偏差的主要因素。羽毛覆盖度会影响鸡体体表温度。通过实验发现,鸡体羽毛覆盖度与体表温度呈极显著负相关(P<0.01),依据羽毛覆盖度对蛋鸡进行分级,对比相同级别、不同区域的蛋鸡体表温度,发现相同级别的羽毛严重损伤区域平均温度比背部平均温度高10℃以上,对比相同区域、不同级别的蛋鸡体表温度,发现背部平均温度最高相差3.89℃,损伤区域平均温度最高相差5.36℃。
In order to solve the problem of insufficient accuracy in evaluating the feather coverage of laying hens by using manual and temperature threshold segmentation methods,thermal infrared images and color images approaches were proposed to calculate the feather coverage.By using Otsu algorithm combined with different color models to extract the chicken body target and the well-covered area of the back feathers and calculate the area ratio,the coverage of the back feathers was obtained.Aiming at the problem of poor segmentation effect caused by the blurred edge of the target chicken body in thermal infrared images,an adaptive thermal infrared image enhancement method based on the gray histogram of R and B components was proposed.The experimental results showed that the segmentation accuracy of chicken body targets and well-covered areas with feathers in thermal infrared images reached 97.18%and 96.86%,and the segmentation accuracy of color images reached 99.58%and 97.86%.The calculation result of the thermal infrared images was closer to the real value of feather coverage than that of color images.The analysis revealed that the roots of white feathers exposed by behaviors of laying hens rubbing against the chicken coop were the main factors for the deviation of the coverage results of the color image calculation.Feather coverage affected chicken body surface temperature.The results showed that the chicken body feather coverage was significantly negatively correlated with the body surface temperature(P<0.01),and the average temperature of the damaged area of the same level was 10℃higher than the average temperature of the back.
作者
杨断利
张然
陈辉
鲍惠玲
宣凤苓
高媛
YANG Duanli;ZHANG Ran;CHEN Hui;BAO Huiling;XUAN Fengling;GAO Yuan(College of Information Science and Technology,Hebei Agricultural University,Baoding 071001,China;Hebei Key Laboratory of Agricultural Big Data,Baoding 071001,China;College of Animal Science and Technology,Hebei Agricultural University,Baoding 071000,China;Shijiazhuang Center of Animal Disease Control,Shijiazhuang 050026,China)
出处
《农业机械学报》
EI
CAS
CSCD
北大核心
2022年第10期242-251,276,共11页
Transactions of the Chinese Society for Agricultural Machinery
基金
国家自然科学基金项目(32172779)
财政部和农业农村部:国家现代农业产业技术体系项目(CARS-40)
关键词
蛋鸡
羽毛覆盖度
体表温度
图像处理
彩色图像
热红外图像
laying hens
feather coverage
body surface temperature
image processing
color image
thermal infrared image
